1. Field of the Invention
The present invention relates to a dishonest registration preventive apparatus for preventing an image of a person other than a concerned person from being dishonestly mixed in images initially registered, in a personal authentication system for collating an image obtained in photographing a person with an image beforehand registered to authenticate the person on the basis of results of the collation.
2. Background Art
Generally known as a personal authentication system is an authentication system based on face collation and described in, for example, JP-A-11-167632. With this system, an image of a concerned person's face is photographed, its feature quantity is beforehand registered, a camera provided at an inlet is used to photograph an image of a concerned person's face at the time of entrance into a specific room to extract its feature quantity, and respective feature quantities of the photographed image and the registered image are compared with each other to discriminate between the concerned person and another person on the basis of a degree of coincidence in feature quantity. Also, with a view to performing authentication more correctly, coincidence in both a face image and an ID (identification number) is also checked by concurrently registering the ID at the time of initial registration of the face image and demanding input of the ID at the time of authentication.
In conventional personal authentication systems, however, when a face image data of another person other than a face image data of a concerned person is included for the same ID at the time of registration of the face image, it will be mistakingly determined at the time of authentication that such another person is the concerned person. Therefore, it is necessary for a supervisor to visually check whether a face image data of another person other than a concerned person is included, which causes an increase in labor cost. Also, while there is also present a method for checking a degree of similarity in a face image data having been registered and determining whether a face image data of another person is included, another person cannot be in some cases identified correctly since it is difficult to identify scatter in the face image data due to orientation of a face and changes in brightness in the case where another person is highly similar to the concerned person.
Further, while an authentication system based on face collation can be introduced into pay content services, or the like provided through a network as Internet propagates in recent years, in which case it is conceivable that in order to have an acquaintance making use of such system, a contractor intermixes a face image of another person to dishonestly register the same at the time of initial registration of a face image. So, since another person making no contract is normally authenticated as the result of collation, such another person can receive a content service or the like at no charge and so a service provider will suffer a loss. Present as a method of finding intermixture and registration of a face image of another person is a method of investigating respective degrees of similarity for registered datum, clustering degrees of similarity, and checking whether two or more clusters are present. Since degrees of similarity disperse even for a concerned person because of orientation of a face, changes in automatic exposure of a camera, or the like, and clusters are divided into sections, it is in some cases difficult to determine presence and absence of intermixture of another person's image from only datum as registered.